SFT-CodeLlama-34B-Instruct-hf_128_v1.1st

This model is a fine-tuned version of meta-llama/CodeLlama-34b-Instruct-hf on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3937

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 2
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Use OptimizerNames.PAGED_ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss
0.7455	0.3049	50	0.7278
0.5717	0.6098	100	0.6302
0.5687	0.9146	150	0.5412
0.4327	1.2195	200	0.4975
0.3597	1.5244	250	0.4572
0.3925	1.8293	300	0.4263
0.262	2.1341	350	0.4220
0.1956	2.4390	400	0.3935
0.17	2.7439	450	0.4036
0.1736	3.0488	500	0.3992
0.1129	3.3537	550	0.4278

Framework versions

• PEFT 0.18.0
• Transformers 4.57.3
• Pytorch 2.9.0+cu126
• Datasets 4.4.2
• Tokenizers 0.22.1